1 / 77

COP 4710: Database Systems Fall 2013 CHAPTERS 16 & 17 – Transaction Processing

COP 4710: Database Systems Fall 2013 CHAPTERS 16 & 17 – Transaction Processing. Instructor : Dr. Mark Llewellyn markl@cs.ucf.edu HEC 236, 407-823-2790 http://www.cs.ucf.edu/courses/cop4710/fall2013. Department of Electrical Engineering and Computer Science

Download Presentation

COP 4710: Database Systems Fall 2013 CHAPTERS 16 & 17 – Transaction Processing

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. COP 4710: Database Systems Fall 2013 CHAPTERS 16 & 17 – Transaction Processing Instructor : Dr. Mark Llewellyn markl@cs.ucf.edu HEC 236, 407-823-2790 http://www.cs.ucf.edu/courses/cop4710/fall2013 Department of Electrical Engineering and Computer Science Computer Science Division University of Central Florida

  2. The execution of any “program” that either accesses (queries) or changes the database contents is called a transaction. Serial transactions – two or more transactions are processed in serial fashion with one transaction starting and completing before the next transaction begins execution. At no time, is more than one transaction processing or making progress. Interleaved transactions – two or more transactions are processed concurrently with only one transaction at a time actually making progress. This most often occurs on a single multi-programmed CPU. Simultaneous transactions – two or more transactions are processed concurrently with any number progressing at one time. This is a multiple CPU situation. Introduction to Transaction Processing

  3. T3 T1 T2 time t0 t1 t2 t3 Serial transactions (unknown number of CPUs) Introduction to Transaction Processing (cont.)

  4. T3 T1 T3 T2 T2 T1 time t0 t1 t2 t3 t4 t5 t6 Interleaved transactions (single CPU) Introduction to Transaction Processing (cont.)

  5. T3 time T2 T1 t0 t1 Simultaneous transactions (3 CPUs shown) Introduction to Transaction Processing (cont.)

  6. When viewed at the transaction level, any transaction has the potential to access the database in two ways: read(item): reads the value of some database item. write(item): write the value of an item into the database. These are not atomic operations. To read an item the following must occur: find the address of the disk block that contains the item. copy the disk block into buffer (if not already present). copy the item from the buffer into the “program”. Introduction to Transaction Processing (cont.)

  7. To write an item the following must occur: find the address of the disk block that contains the item. copy the disk block into buffer (if not already present). copy the item from the buffer into the “program”. store the updated block from the buffer back onto the disk (at some point in time, usually not immediately). When to write back is typically up to the recovery system of the database and may involve OS control. Too early of a write back may cause unnecessary data transfers. Too late of a write back may cause unnecessary blocking. Introduction to Transaction Processing (cont.)

  8. Given a consistent (correct?) state of the database as input an individually correct transaction will produce a correct state of the database as output, if that transaction is executed in isolation. The goal of concurrency control is to allow multiple transactions to be processing simultaneously within a certain time period with all of the concurrent transactions producing a correct state of the database at then end of their concurrent execution. Concurrency Control

  9. There are many different types of conflicts that can occur between concurrently executing processes if concurrency control is not enforced. Lost Update Problem (A Write-Write Conflict) (overwriting uncommitted data) Suppose two distinct transactions T1 and T2 are processing in the concurrent order shown below accessing a common value n. timeactioncomment t0 T1 performs read(n) suppose T1 reads value of n = 5 t1 T2 performs read(n) T2 will read a value of n = 5 t2 T1 performs write(n-1) T1 will write value of n = 4 t3 T2 performs write(n-1) T2 will also write value of n = 4! Problem: The update performed by T1 at time t2 is “lost” since the update written by T2 at time t3 overwrites the previous value. Concurrency Control – Why Its Needed

  10. There are several different ways in which the lost update problem can be handled. Prevent T2 from reading the value of n at time t1 on the grounds that T1 has already read the value of n and may therefore update the value. Prevent T1 from writing the value of n-1 at time t2 on the grounds that T2 has also read the same value of n and would therefore be executing on an obsolete value of n, since T2 cannot re-read n. Prevent T2 from writing the value of n-1 at time t3 on the grounds that T1 has already updated the value of n and since T1 preceded T2, then T2 is using an obsolete value of n. The first two of these techniques can be implemented using locking protocols, while the third technique can be implemented with time-stamping. We’ll see both of these techniques later. Handling the Lost Update Problem

  11. Dirty Read Problem (A Write-Read Conflict)(reading uncommitted data) Suppose two distinct transactions T1 and T2 are processing in the concurrent order shown below accessing a common value n. timeactioncomment t0 T1 performs read(n) suppose T1 reads value of n = 5 t1 T1 performs write(n-1) T1 writes a value of n = 4 t2 T2 performs read(n) T2 will read value of n = 4 t3 T1 aborts T2 is executing with a “bad” value of n Problem: T2 is operating with a value that was written by a transaction that aborted prior to the completion of T2. When T1 aborts all of its updates must be undone, which means that T2 is executing with a bad value of n and therefore cannot leave the database in a consistent state. Solution: T2 must also be aborted. The Dirty Read Problem

  12. Unrepeatable Read Problem (A Read-Write Conflict) Suppose two distinct transactions T1 and T2 are processing in the concurrent order shown below accessing a common value n. timeactioncomment t0 T1 performs read(n) suppose T1 reads value of n = 5 t1 T1 performs read(n) T1 reads a value of n = 5 t2 T2 performs write(n-1) T2 will write value of n = 4 t3 T1 performs read(n) T1 reads a different value of n this time Problem: When T1 performs its second read of n, the value is not the same as its first read of n. T1 cannot repeat its read. Solution: This problem is typically handled with locking which is rather inflexible, but can also be solved with time-stamping. The Unrepeatable Read Problem

  13. Whenever a transaction is submitted to the DBMS for execution, the DBMS is responsible for making sure that either: All operations of the transaction are completed successfully and their effect is permanently recorded in the database, or The transaction has no effect whatsoever on the database or any other transaction. If a transaction fails after executing some of its operations, problems will occur with consistency in the database. Therefore, if a transaction fails after its is initiated but prior to its commitment, all of the effects of that transaction must be undone from the database. The Transaction Recovery System

  14. Types of failures for a transaction: System crash – some type of hardware or system failure occurs. Transaction error – integer overflow, division by zero, operator intervention. Local errors or exception conditions – required data is not available. Concurrency control enforcement – serializability is violated, deadlock detection victim selection, etc. Disk errors – error correction/detection. Physical problems – fire, power failure, operator error, etc. The Transaction Recovery System (cont.)

  15. A transaction can be in one of several different states: begin_transaction: marks the beginning of the transaction. read/write: specifies the various db operations performed by the transaction. end_transaction: specifies that all read/write operations have ended and the transaction is ready to terminate. Note: this does not actually end the transactions time in the system – now it heads to the concurrency control system for verification. commit: marks the successful end of the transaction – its effects are now permanent (committed) in the database and cannot be undone. abort (rollback): marks the unsuccessful end of the transaction. All changes and effects in the database must be undone and/or other transactions must be aborted. No changes are committed for the transaction. The States of a Transaction

  16. begin_transaction committed commit partially committed end_transaction active read/write abort abort failed terminated The States of a Transaction (cont.)

  17. The system log keeps track of all transaction operations that affect values of database items. The information in the log is used to perform recovery operations from transaction failures. Most logs consist of several levels ranging from the log maintained in main memory to archival versions on backup storage devices. Upon entering the system, each transaction is given a unique transaction identifier (timestamps are common). System Log

  18. In the system log, several different types of entries occur depending on the action of the transaction: [start, T]: begin transaction T. [write, T, X, old, new]: transaction T performs a write on object X, both old and new values of X are recorded in the log entry. [read, T, X]: transaction T performs a read on object X. [commit, T]: transaction T has successfully completed and indicates that its changes can be made permanent. [abort, T]: transaction T has aborted. Some types of recovery protocols do not require read operations be logged. System Log (cont.)

  19. A transaction T reaches its commit point when all of its operations that access the database have successfully completed and the effect of all of these operations have been recorded in the log. Beyond the commit point, a transaction is said to be committed and its effect on the database is assumed to be permanent. It is at this point that [commit, T] is entered into the system log. If a failure occurs, a search backward through the log (in terms of time) is made for all transactions that have written a [start, T] into the log but have not yet written [commit, T] into the log. This set of transactions must be rolled back. Commit Point

  20. Atomcity – a transaction is an atomic unit of processing; it is either performed in its entirety or not at all. Consistency – a correct execution of the transaction must take the database from one consistent state to another. Isolation – a transaction should not make its updates visible to other transactions until it is committed. Strict enforcement of this property solves the dirty read problem and prevents cascading rollbacks from occurring. Durability – once a transaction changes the database and those changes are committed, the changes must never be lost because of a failure. ACID Properties of Transactions

  21. When transactions are executing concurrently in an interleaved fashion, the order of execution of the operations from the various transactions forms what is known as a transaction schedule (sometimes called a history). Schedules and Recoverability A schedule S of n transactions T1, T2, T3, ..., Tn is an ordering of the operations of the transactions where for each transaction Ti S, each operation in Ti occurs in the same order in both Ti and S.

  22. The notation used for depicting schedules is: ri(x) means that transaction i performs a read of object x. wi(x) means that transaction i performs a write of object x. ci means that transaction i commits. ai means that transaction i aborts. An example schedule: SA = (r1(x), r2(x), w1(x), w2(x), c1, c2) This example schedule represents the lost update problem. Another example: SB = (r1(x), r1(y), w1(y), r2(x), w1(x), w2(y), c2, c1) Schedules and Recoverability (cont.)

  23. Two operations in a schedule are said to conflict if they belong to different transactions, access the same item, and one of the operations is a write operation. Consider the following schedule: SA = (r1(x), r2(x), w1(x), c1, c2) r2(x) and w1(x) conflict r1(x) and r2(x) do not conflict. Conflict in a Schedule

  24. For some schedules it is easy to recover from transaction failures, while for others it can be quite difficult and involved. Recoverability from failures depends in large part on the scheduling protocols used. A protocol which never rolls back a transaction once it is committed is said to be a recoverable schedule. Within a schedule a transaction T is said to have read from a transaction T* if in the schedule some item X is first written by T* and subsequently read by T. Recoverability

  25. A schedule S is a recoverable schedule if no transaction T in S commits until all transactions T* that have written an item which T reads have committed. For each pair of transactions Tx and Ty, if Ty reads an item previously written by Tx, then Tx must commit before Ty. Example: SA = (r1(x), r2(x), w1(x), r1(y), w2(x), c2, w1(y), c1) This is a recoverable schedule since, T2 does not read any item written by T1 and T1 does not read any item written by T2. Example: SB = (r1(x), w1(x), r2(x), r1(y), w2(x), c2, a1) This is not a recoverable schedule since T2 reads value of x written by T1 and T2 commits before T1 aborts. Since T1 aborts, the value of x written by T2 must be invalid so T2 which has committed must be rolled back rendering schedule SB not recoverable. Recoverability (cont.)

  26. Cascading rollback occurs when an uncommitted transaction must be rolled back due to its read of an item written by a transaction that has failed. Example: SA = (r1(x), w1(x), r2(x), r1(y), r3(x), w2(x), w1(y), a1) In SA, T3 must be rolled back since T3 read value of x produced by T1 and T1 subsequently failed. T2 must also be rolled back since T2 read value of x produced by T1 and T1 subsequently failed. Example: SB = (r1(x), w1(x), r2(x), w2(x), r3(x), w1(y), a1) In SB, T2 must be rolled back since T2 read value of x produced by T1 and T1 subsequently failed. T3 must also be rolled back since T3 read value of x produced by T2 and T2 subsequently failed. T3 is rolled back, not because of the failure of T1 but because of the failure of T2. Cascading Rollback

  27. Cascading rollback can be avoided in a schedule if every transaction in the schedule only reads items that were written by committed transactions. A strict schedule is a schedule in which no transaction can read or write an item x until the last transaction that wrote x has committed (or aborted). Example: SA = (r1(x), w1(x), c1, r2(x), c2) Cascading Rollback (cont.)

  28. Given two transactions T1 and T2, if no interleaving of the transactions is allowed (they are executed in isolation), then there are only two ways of ordering the operations of the two transactions. Either: (1) T1 executes followed by T2 or (2) T2 executes followed by T1 Interleaving of the operations of the transactions allows for many possible orders in which the operations can be performed. Serializability

  29. Serializability theory determines which schedules are correct and which are not and develops techniques which allow for only correct schedules to be executed. Interleaved execution, regardless of what order is selected, must have the same effect of some serial ordering of the transactions in a schedule. A serial schedule is one in which every transaction T that participates in the schedule, all of the operations of T are executed consecutively in the schedule, otherwise the schedule is non-serial. Serializability (cont.)

  30. A concurrent (or interleaved) schedule of n transactions is serializable if it is equivalent (produces the same result) to some serial schedule of the same n transactions. A schedule of n transactions will have n! serial schedules and many more non-serial schedules. Example: Transactions T1, T2, and T3 have the following serial schedules: (T1, T2, T3), (T1, T3, T2), (T2, T1, T3), (T2, T3, T1), (T3, T1, T2), and (T3, T2, T1). There are two disjoint sets of non-serializable schedules: Serializable: those non-serial schedules which are equivalent to one or more of the serial schedules. Non-serializable: those non-serial schedules which are not equivalent to any serial schedule. Serializability (cont.)

  31. There are two main types of serializable schedules: Conflict serializable: In general this is an O(n3) problem where n represents the number of vertices in a graph representing distinct transactions. View serializable: This is an NP-C problem, meaning that the only known algorithms to solve it are exponential in the number of transactions in the schedule. We’ll look only a conflict serializable schedules. Recall that two operations in a schedule conflict if (1) they belong to different transactions, (2) they access the same database item, and (3) one of the operations is a write. Serializability (cont.)

  32. If the two conflicting operations are applied in different orders in two different schedules, the effect of the schedules can be different on either the transaction or the database, and thus, the two schedules are not conflict equivalent. Example: SA = (r1(x), w2(x)) SB = (w2(x), r1(x)) The value of x read in SA may be different than in SB. Example: SA = (w1(x), w2(x), r3(x)) SB = (w2(x), w1(x), r3(x)) The value of x read by T3 may be different in SA than in SB Conflict Serializability

  33. To generate a conflict serializable schedule equivalent to some serial schedule using the notion of conflict equivalence involves the reordering of non-conflicting operations of the schedule until an equivalent serial schedule is produced. The technique is this: build a precedence graph based upon the concurrent schedule. Use a cycle detection algorithm on the graph. If a cycle exists, S is not conflict serializable. If no cycle exists, a topological sort of the graph will yield an equivalent serial schedule. Conflict Serializability (cont.)

  34. Algorithm Conflict_Serializable • Algorithm Conflict_Serializable • //input: a concurrent schedule S • //output: no – if S is not conflict serializable, a serial schedule S* equivalent to S otherwise. • Conflict_Serializable(S) • for each transaction TX S, create a node (in the graph) labeled TX. • (RAW: READ AFTER WRITE) for each case in S where TY executes read(a) after TX executes write(a) create the edge TX TY. The meaning of this edge is that TX must precede TY in any serially equivalent schedule. • (WAR: WRITE AFTER READ) for each case in S where TY executes write(a) after TX executes read(a) create the edge TX  TY. The meaning of this edge is that TX must precede TY in any serially equivalent schedule. • (WAW: WRITE AFTER WRITE) for each case in S where TY executes write(a) after TX executes write(a) create the edge TX  TY. The meaning of this edge is that TX must precede TY in any serially equivalent schedule. • if the graph contains a cycle then return no, otherwise topologically sort the graph and return a serial schedule S* which is equivalent to the concurrent schedule S.

  35. Conflict Serializability – Example #1 Let SC = (r1(a), w1(a), r2(a) w2(a), r1(b), w1(b)) 1: r1(a) precedes w2(a) (WAR) 2: w1(a) precedes r2(a) (RAW) 3: w1(a) precedes w2(a) (WAW) T2 T1 1,2,3 Graph contains no cycle, so SC is conflict serializable

  36. Conflict Serializability – Example #1 Let SC = (r1(a), r2(a), w1(a), r1(b), w2(a), w1(b)) w1(a) precedes r2(a) (RAW) 1,3 1: r1(a) precedes w2(a) (WAR) 2: r2(a) precedes w1(a) (WAR) 3: w1(a) precedes w2(a) (WAW) T2 T1 2 Graph contains a cycle, so SC is not conflict serializable

  37. Let SC = (r3(y), r3(z), r1(x), w1(x), w3(y), w3(z), r2(z), r1(y), w1(y), r2(y), w2(y), r2(y), w2(y) ) T2 T1 T3 Conflict Serializability – Example #2 • edgereason • w3(y) precedes r2(y) (RAW) • w1(x) precedes r2(x) (RAW) • w3(z) precedes r2(z) (RAW) • w1(y) precedes r2(y) (RAW) • r3(y) precedes w1(y) (WAR) • r1(x) precedes w2(x) (WAR) • r1(y) precedes w2(y) (WAR) • There are seven other conflicts that can be found in this schedule, but none of them will introduce a cycle. Find the missing seven. 2, 4, 6, 7 5 1, 3 Graph contains no cycles, so a serially equivalent schedule would be T3, T1, T2.

  38. There are several different techniques that can be employed to handle concurrent transactions. The basic techniques fall into one of four categories: Locking protocols Timestamping protocols Multiversion protocols – deal with multiple versions of the same data Optimistic protocols – validation and certification techniques Concurrency Control Techniques

  39. Transactions “request” locks and “release” locks on database objects through a system component called a lock manager. Main issues in locking are: What type of locks are to be maintained. Lock granularity: runs from very coarse to very fine. Locking protocol Deadlock, livelock, starvation Other issues such as serializability Locking Protocols grant issue lock – transaction continues LOCK MANAGER process request abort deny block in queue

  40. Locking protocols are quite varied in their degree of complexity and sophistication, ranging from very simple yet highly restrictive protocols, to quite complex protocols which nearly rival time-stamping protocols in their flexibility for allowing concurrent execution. In order to give you a flavor of how locking protocols work, we’ll focus on only the most simple locking protocols. While the basic techniques of all locking protocols are the same, in general, the more complex the locking protocol the higher the degree of concurrent execution that will be permitted under the protocol. Locking Protocols (cont.)

  41. When devising a locking protocol, one of the first things that must be considered is the level of locking that will be supported by the protocol. Simple protocols will support only a single level of locking while more sophisticated protocols can support several different levels of locking. The locking level (also called the locking granularity), defines the type of database object on which a lock can be obtained. The coarsest level of locking is at the database level, a transaction basically locks the entire database while it is executing. Serializability is ensured because with the entire database locked, only one transaction can be executing at a time, which ensures a serial schedule of the transactions. Locking Granularity

  42. Moving toward a finer locking level, typically the next level of locking that is available is at the relation (table) level. In this case, a lock is obtained on each relation that is required by a transaction to complete its task. If we have two transactions which need different relations to accomplish their tasks, then they can execute concurrently by obtaining locks on their respective relations without interfering with one another. Thus, the finer grain lock has the potential to enhance the level of concurrency in the system. The next level of locking is usually at the tuple level. In this case several transactions can be executing on the same relation simultaneously, provided that they do not need the same tuples to perform their tasks. At the extreme fine end of the locking granularity would be locks at the attribute level. This would allow multiple transactions to be simultaneously executing in the same relation in the same tuple, as long as they didn’t need the same attribute from the same tuple at the same time. At this level of locking the highest degree of concurrency will be achieved. Locking Granularity (cont.)

  43. There is, unfortunately a trade-off between enhancing the level of concurrency in the system and the ability to manage the locks. At the coarse end of the scale we need to manage only a single lock, which is easy to do, but this also gives us the least degree of concurrency. At the extremely fine end of the scale we would need to manage an extremely large number of locks in order to achieve the highest degree of concurrency in the system. Unfortunately, with VLDB (Very Large Data Bases) the number of locks that would need to be managed at the attribute level poses too complex of a problem to handle efficiently and locking at this level almost never occurs. Locking Granularity (cont.)

  44. For example, consider a fairly small database consisting of 10 relations each with 10 attributes and suppose that each relation has 1000 tuples. This database would require the management of 10  10  1000 = 100,000 locks. A large database with 50 relations each having 25 attributes and assuming that each relation contained on the order of a 100,000 tuples; the number of locks that need to be managed grows to 1.25108 (125 million locks). A VLDB with hundreds of relations and hundreds of attributes and potentially millions of tuples can easily require billions of locks to be maintained if the locking level is at the attribute level. Due to the potentially overwhelming number of locks that would need to be maintained at this level, a compromise to the tuple level of locking is often utilized. Locking Granularity (cont.)

  45. There are different types of locks that locking protocols may utilize. The most restrictive systems use only exclusive-locks (X-lock also called a binary lock). An exclusive lock permits the transaction which holds the lock exclusive access to the object of the lock. The process of locking and un-locking objects must be indivisible operations within a critical section. There can be no interleaving of issuing and releasing locks. Types of Locks If transaction TX holds an X-lock on object A then no distinct transaction TY can obtain an X-lock on object A until transaction TX releases the X-lock on object A. TY is blocked awaiting the X-lock on object A.

  46. When the lock manager grants a transaction’s request for a particular lock, the transaction is said to “hold the lock” on the object. Under the X-lock protocol a transaction must obtain, for every object required by the transaction, an X-lock on the object. This applies to both reading and writing operations. X-Lock Protocol Before any transaction TX can read or write an object A, it must first acquire an X-lock on object A. If the request is granted TX will proceed with execution. If the request is denied, TX will be placed into a queue of transactions awaiting the X-lock on object A, until the lock can be granted. After TX finishes with object A, it must release the X-lock.

  47. Serializability Under X-Lock Protocol • Algorithm TestSerializiabiltyXLock • //input: a concurrent schedule S under X-lock protocol • //output: if S is serializable, then a serially equivalent schedule S is produced, otherwise, no. • TestSerializabilityXLock(S) • let S = (a1, a2, ..., an) where “action” ai is either (TX: Xlock A) or (TX: Unlock A) • construct a precedence graph of n nodes where n is the number of distinct transactions in S. • proceed through S as follows: • if ar = (TX: Unlock A) then look for the next action as of the form (TY: Xlock A). If one exists, draw an edge in the graph from TX to TY. The meaning of this edge is that in any serially equivalent schedule TX must precede TY. • if the graph constructed in step 3 contains a cycle, then S is not equivalent to any serial schedule (i.e., S is not serializable). If no cycle exists, then any topological sort of the graph will yield a serial schedule equivalent to S.

  48. Let S = [(T1: Xlock A), (T2: Xlock B), (T2: Xlock C), (T2: Unlock B), (T1: Xlock B), (T1: Unlock A), (T2: Xlock A), (T2:Unlock C), (T2: Unlock A), (T3: Xlock A), (T3: Xlock C), (T1: Unlock B), (T3: Unlock C), (T3: Unlock A)] Example - X-Lock Protocol and Serializability 2 Edge #1: (T2: Unlock B)...(T1:Xlock B) Edge #2: (T1: Unlock A)...(T2: Xlock A) Edge #3: (T2: Unlock C)...(T3: Xlock C) Edge #4: (T2: Unlock A)...(T3: Xlock A) T1 T2 1 3 , 4 T3 Not serializable, cycle exists

  49. The X-lock protocol is too restrictive. Several transactions that need only to read an object must all wait in turn to gain an X-lock on the object, which unnecessarily delays each of the transactions. One solution is to issue different types of locks, called shared-locks (S-locks or read-locks) and write-locks (X-locks). The lock manager can grant any number of shared locks to concurrent transactions that need only to read an object, so multiple reading is possible. Exclusive locks are issued to transactions needing to write an object. If an X-lock has been issued on an object to transaction TX, then no other distinct transaction TY can be granted either an S-lock or an X-lock until TX releases the X-lock. If any transaction TX holds an S-lock on an object, then no other distinct transaction TY can be granted an X-lock on the object until all S-locks have been released. Problems with X-Lock Protocol

  50. Serializability Under X/S-Lock Protocol • Algorithm TestSerializiabiltyX/SLock • //input: a concurrent schedule S under X/S-lock protocol • //output: if S is serializable, then a serially equivalent schedule S is produced, otherwise, no. • TestSerializabilityXLock(S) • let S = (a1, a2, ..., an) where “action” ai is one of (TX: Slock A), (TX: Xlock A) or (TX: Unlock A). • construct a precedence graph of n nodes where n is the number of distinct transactions in S. • proceed through S as follows: • if ax = (TX: Slock A) and ay is the next action (if it exists) of the form (TY: Xlock A) then draw an edge from TX to TY. • if ax = (TX: Xlock A) and there exists an action az = (TZ: Xlock A) then draw an edge in the graph from TX to TZ. Also, for each action ay of the form (TY: Slock A) where ay occurs after ax (TX: Unlock A) but before aZ (TZ: Xlock A) draw an edge from TX to TY. If az does not exist, then TY is any transaction to perform (TY: Slock A) after (TX: Unlock A). • if the graph constructed in step 3 contains a cycle, then S is not equivalent to any serial schedule (i.e., S is not serializable). If no cycle exists, then any topological sort of the graph will yield a serial schedule equivalent to S.

More Related